Overview

What I’ve been doing is based on datasets which are already filtered to exclude any data impacted by stray light, lightning, lunar illumination, and cloud-cover. It’s provided by NOAA(https://www.ngdc.noaa.gov/eog/viirs/).However, it’s recorded every month and it has some missing values so it may not be as accurate as the daily data.
Time: I chose 2014 and 2017 to do the analyze because the Ramadans in these two years are very close to the natural periods related to the motion of the Moon.
Places: Antalya,Sanliurfa(big cities with intensive Muslim places of worship) and Siirt(big city with sparse Muslim places of worship).
Besides, I also compared the NTL at July from 2013-2017 to see if there is any significant increase in 2014 since July and Ramadan overlapped only in that year.

Data visulization

28 large Turkey cities from west to east are mapped and colored in order to reveal the gradations in each local urban environment in the Ramadan of 2017.
Within each city, the relative intensity of light is indicative of the differences in human activity: brighter yellow indicates relatively more population acitivity and darker blue indicates relatively less activity. Also, the development patterns are distinct in each city, contoured to the waterways and roadways.
As we can see in the map, central business districts such as Ankara are appear in brilliant yellow clusters.

Turkey_metrocities <-c(
  #west
  "Eskisehir","Izmir","Manisa","Aydin","Denizli","Antalya","Balikesir",
  "Mugla","Canakkale", 
  
  #middle
  "Adana",
  "Ankara","Samsun","Malatya","Kayseri","Kahramanmaras", "Gaziantep",
  "Hatay",  "Mersin",
  
  #east
  "Trabzon","Ordu","Ardahan","Igdir","Sirnak","Mardin","Erzurum","Konya",
  "Diyarbakir", "Sanliurfa")

##Set graph layout
windows.options(width = 10, height = 8)
par(mai=c(0,0,0,0),mfrow = c(7,4),bg='#001a4d', bty='n')

##Loop through data
coords <- data.frame() ##place holder
for(i in 1:length(Turkey_metrocities)){
  
  ##Coords
  temp_coord <- geocode(Turkey_metrocities[i], source = "google")
  coords <- rbind(coords,temp_coord)
  range <- extent(temp_coord$lon - 1, temp_coord$lon + 1,
                  temp_coord$lat - 0.25, temp_coord$lat + 0.25)
  rc <- crop(rast_201706, range)    
  
  ##Rescale brackets
  sampled <- as.vector(rc)
  if (sum(sampled)==0){
    next
  }
  clusters <- 15
  clust <- kmeans(sampled,clusters)$cluster
  brk <- sort(aggregate(sampled, list(clust), max)[,2])
  
  #Plots
  plot(rc, breaks=brk,     
       col=colorRampPalette(c("#001a4d","#0066FF","yellow"))(clusters), 
       legend=F,yaxt='n',xaxt='n',frame = F, asp=1.5)
  text(temp_coord$lon ,temp_coord$lat + 0.15,
       Turkey_metrocities[i], 
       col="white", cex=1.25)
}

Radiance and population

In direct comparison, total night-time radiance and population are positively correlated. More population could cause more human activities and then results in stronger radiance.

#population vs NTL
Pops <- read.csv(file="F:/Lingyi/population.csv", header=TRUE)

rads = data.frame()
msa_list <- c(1:81)
for (i in msa_list){
  #print(i)
  shp_temp <- turkeyshape[turkeyshape@data$ID_1==i,]
  loc = shp_temp@data$NAME_1
  rad <- masq(shp_temp,rast_201706,1)$avg_rad
  temp = data.frame(asciiname = loc, rad_sum = sum(rad))
  rads <-rbind(rads,temp)
}

total <- merge(rads,Pops, by = "asciiname")
colnames(total) <- c("asciiname" , "rad_sum"  ,  "population" ,"location" )

plot_ly(total,x = ~log(rad_sum), y = ~log(population),
        text = paste("City:", total$asciiname),
        mode = "markers",
        color = total$rad_sum, colors="Spectral")  %>% 
  layout(title="Total Nighttime radiance vs. Population", showlegend = F)

Muslim places of worship

The Muslim places of worship dataset comes from the OpenStreetMap project which are normally updated everyday.(http://download.geofabrik.de/)
As we can see in the plot, the Muslim places of worships gather in Istanbul, Ankara, Antalya, Sanliurfa,Kayseri, Ordu and etc.
Therefore, I chose cities as follows to do the analysis:
* Antalya (big city with intensive Muslim places of worship)
* Sanliurfa(big city with intensive Muslim places of worship)
* Siirt(big city with sparse Muslim places of worship)

Method

Firstly, I used the turkey shapefile(from https://stacks.stanford.edu/file) to get the exact extent of a specific city and then get the subdataset.
Then I divided the subdataset into approximate 300,000 pixels based on the size of the area. Each pixel corresponding to a number(radiance) with its longitude and latitude.
After that, I divided radiance into several levels and plot histograms to show the counts of each radiance interval and calculate the total night-time radiance.

2017 data explore

The 2017 Ramadan started from May 27 and ended in June 24, So I used NTL datasets from May, June and July to see the difference of night-time radiance over these three months in three cities.

Antalya

Antalya is the eighth-most populous city in Turkey and the capital of Antalya Province. Located on Anatolia’s southwest coast bordered by the Taurus Mountains, Antalya is the largest Turkish city on the Mediterranean coast with over one million people in its metropolitan area.

In the plot above, the TNL(total night-time lights) at Antalya was not increased at June as we imagine and the radiance distribution was the same as usual.

Sanliurfa

Sanliurfa is a city with a population of over 2 million residents in south-eastern Turkey, and the capital of Sanlıurfa Province. Danliurfa is a multiethnic city with a Turkish, Kurdish, Armenian and Arab population and is situated on a plain about eighty kilometres east of the Euphrates River.

The TNL at Sanliurfa increased a little at Ramadan and was more evenly distributed.

Siirt

Siirt is a city in southeastern Turkey and the seat of Siirt Province. The population of the city according to the 2009 census was 129,188. The majority of the city’s population is Kurdish.

The TNL at Sanliurfa was not increased as we imagine and in fact it was decreased a little bit.

2014 data explore

The 2014 Ramadan started from June 28 and ended in July 27. So I was supposed to use NTL datasets from June, July and August. However, when I read in the data in June, it turned out to be all zeros. Therefore, I used the data in May instead which was two months before the Ramadan.

Antalya

In the plot above we can see an increasing of TNL at Ramadan and the count at the level which log(radiance)=-1.5 was extramly high. That means a lot of places at Antalya was remain a below average brightness at Ramadan.

Sanliurfa

There was also an significant increasing of TNL at Ramadan in Sanliurfa. But the distribution pattern was a bit different as usual. The counts in the lower radiance levels were increased which means that the areas where lights were rarely turned on at night was getting some human activities at Ramadan.

Siirt

There was also an significant increasing of TNL at Ramadan in Siirt, though there may not be as much Muslim as other two places. Radiance was centred around -1.5 to 0. We can also see that the lower level radiances disappeared at the month after the Ramadan.

NTL at July from 2013 to 2017

Since July and Ramadan overlapped only in 2014, we could compare the night-time light radiance of July from 2013 to 2017 to demonstrate the effect of Ramadan.

We do see an significant inceasing of TNL at 2014 in Antalya which may indicate more human activities at Ramadan. From the distribution pattern we can see that lowers radiances levels disappeared and these pixels were added to ralatively higher levels(-2.5 to -1).

The radiance in 2017 appereared to be extraordinary high so we may not consider it when doing the comparition.

At the city with sparse places of worships, we could not see any important changes at July 2014.

I believe there are some missing values in the dataset of 2015.